CN117471346A - Method and system for determining remaining life and health status of retired battery module - Google Patents
Method and system for determining remaining life and health status of retired battery module Download PDFInfo
- Publication number
- CN117471346A CN117471346A CN202311465615.7A CN202311465615A CN117471346A CN 117471346 A CN117471346 A CN 117471346A CN 202311465615 A CN202311465615 A CN 202311465615A CN 117471346 A CN117471346 A CN 117471346A
- Authority
- CN
- China
- Prior art keywords
- data
- module
- health
- model
- life
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
- 238000000034 method Methods 0.000 title claims abstract description 23
- 230000003862 health status Effects 0.000 title claims description 30
- 230000036541 health Effects 0.000 claims abstract description 76
- 238000011156 evaluation Methods 0.000 claims abstract description 19
- 238000007781 pre-processing Methods 0.000 claims abstract description 19
- 238000000605 extraction Methods 0.000 claims abstract description 17
- 238000012549 training Methods 0.000 claims description 24
- 230000006870 function Effects 0.000 claims description 12
- 210000002569 neuron Anatomy 0.000 claims description 12
- 230000008859 change Effects 0.000 claims description 11
- 238000013136 deep learning model Methods 0.000 claims description 10
- 238000003062 neural network model Methods 0.000 claims description 9
- 238000013528 artificial neural network Methods 0.000 claims description 8
- 230000008569 process Effects 0.000 claims description 8
- 230000004913 activation Effects 0.000 claims description 6
- 239000011159 matrix material Substances 0.000 claims description 6
- 238000012544 monitoring process Methods 0.000 claims description 6
- 230000002159 abnormal effect Effects 0.000 claims description 5
- 125000004122 cyclic group Chemical group 0.000 claims description 3
- 238000013135 deep learning Methods 0.000 claims description 3
- 239000000284 extract Substances 0.000 claims description 3
- 238000009499 grossing Methods 0.000 claims description 3
- 230000001771 impaired effect Effects 0.000 claims description 3
- 238000013507 mapping Methods 0.000 claims description 3
- 238000013178 mathematical model Methods 0.000 claims description 3
- 238000010606 normalization Methods 0.000 claims description 3
- 230000009467 reduction Effects 0.000 claims description 3
- 238000012360 testing method Methods 0.000 claims description 3
- 238000010200 validation analysis Methods 0.000 claims description 3
- 238000004364 calculation method Methods 0.000 claims description 2
- 238000001514 detection method Methods 0.000 abstract description 10
- 238000012423 maintenance Methods 0.000 abstract description 4
- 239000010926 waste battery Substances 0.000 abstract description 2
- 230000009286 beneficial effect Effects 0.000 description 3
- 238000013461 design Methods 0.000 description 3
- 230000007613 environmental effect Effects 0.000 description 2
- WHXSMMKQMYFTQS-UHFFFAOYSA-N Lithium Chemical compound [Li] WHXSMMKQMYFTQS-UHFFFAOYSA-N 0.000 description 1
- HBBGRARXTFLTSG-UHFFFAOYSA-N Lithium ion Chemical compound [Li+] HBBGRARXTFLTSG-UHFFFAOYSA-N 0.000 description 1
- 239000002253 acid Substances 0.000 description 1
- 230000009471 action Effects 0.000 description 1
- 230000004075 alteration Effects 0.000 description 1
- 230000006399 behavior Effects 0.000 description 1
- 238000013145 classification model Methods 0.000 description 1
- 230000007547 defect Effects 0.000 description 1
- 238000004146 energy storage Methods 0.000 description 1
- 229910052744 lithium Inorganic materials 0.000 description 1
- 229910001416 lithium ion Inorganic materials 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 238000006467 substitution reaction Methods 0.000 description 1
- 239000002699 waste material Substances 0.000 description 1
Classifications
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01R—MEASURING ELECTRIC VARIABLES; MEASURING MAGNETIC VARIABLES
- G01R31/00—Arrangements for testing electric properties; Arrangements for locating electric faults; Arrangements for electrical testing characterised by what is being tested not provided for elsewhere
- G01R31/36—Arrangements for testing, measuring or monitoring the electrical condition of accumulators or electric batteries, e.g. capacity or state of charge [SoC]
- G01R31/392—Determining battery ageing or deterioration, e.g. state of health
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01R—MEASURING ELECTRIC VARIABLES; MEASURING MAGNETIC VARIABLES
- G01R31/00—Arrangements for testing electric properties; Arrangements for locating electric faults; Arrangements for electrical testing characterised by what is being tested not provided for elsewhere
- G01R31/36—Arrangements for testing, measuring or monitoring the electrical condition of accumulators or electric batteries, e.g. capacity or state of charge [SoC]
- G01R31/367—Software therefor, e.g. for battery testing using modelling or look-up tables
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01R—MEASURING ELECTRIC VARIABLES; MEASURING MAGNETIC VARIABLES
- G01R31/00—Arrangements for testing electric properties; Arrangements for locating electric faults; Arrangements for electrical testing characterised by what is being tested not provided for elsewhere
- G01R31/36—Arrangements for testing, measuring or monitoring the electrical condition of accumulators or electric batteries, e.g. capacity or state of charge [SoC]
- G01R31/382—Arrangements for monitoring battery or accumulator variables, e.g. SoC
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01R—MEASURING ELECTRIC VARIABLES; MEASURING MAGNETIC VARIABLES
- G01R31/00—Arrangements for testing electric properties; Arrangements for locating electric faults; Arrangements for electrical testing characterised by what is being tested not provided for elsewhere
- G01R31/36—Arrangements for testing, measuring or monitoring the electrical condition of accumulators or electric batteries, e.g. capacity or state of charge [SoC]
- G01R31/396—Acquisition or processing of data for testing or for monitoring individual cells or groups of cells within a battery
Landscapes
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Secondary Cells (AREA)
Abstract
The invention provides a method and a system for determining the residual life and the health state of a retired battery module, and relates to the technical field of waste battery detection. The method and the system for the rest life and the health state of the retired battery module comprise a data acquisition module, a data preprocessing module, a characteristic extraction module, a rest life model building module, a rest life prediction module and a result output and health state evaluation module; the data acquisition module comprises a current sensor, a voltage sensor and a temperature sensor; the data preprocessing module cleans and preprocesses the collected original data. Through repeated iterative detection, more data samples can be accumulated, a more accurate residual life prediction model is facilitated to be established, and by utilizing the data, a worker can more accurately predict the life of the battery module and take corresponding maintenance and management measures in advance.
Description
Technical Field
The invention relates to the technical field of waste battery detection, in particular to a method and a system for determining the residual life and the health state of a retired battery module.
Background
In recent years, lithium ion batteries have been commercially used in various fields on a large scale as new energy storage devices with long life and environmental protection, particularly in the field of electric automobiles, and have now become the mainstream in place of conventional lead-acid batteries. With the growth of the lithium battery industry, battery management systems are rapidly developed, and a system for retired battery modules to prolong the service life and state of health is usually a software or hardware system for monitoring, evaluating and predicting the service life and state of health of the battery modules, and by monitoring the service life and state of health, some measures can be taken to prolong the service life of the battery, for example, optimizing a charge and discharge strategy according to the actual state of the battery, avoiding excessive charge and discharge, reducing damage to the battery and prolonging the service life of the battery.
The current prediction parameter value of the residual life of the battery module can not directly reflect the actual health state of the battery under the current detection turntable, the health state of the battery is affected by various factors including charge and discharge history, environmental conditions, use modes and the like, the parameters obtained by only relying on parameter identification can not comprehensively and accurately reflect the influence of the factors on the health of the battery, the detection accuracy is low, the residual resource waste of the retired battery is easy to cause, the performance change trend of the retired battery module can not be accurately reflected, the residual life and the health state of the retired battery module are difficult to accurately evaluate only by one detection result, the problems of capacity attenuation, internal resistance increase and temperature rise can not be timely found by staff, and a certain degree of potential safety hazard is easy to cause.
Disclosure of Invention
(one) solving the technical problems
Aiming at the defects of the prior art, the invention provides a method and a system for determining the residual life and the health state of the retired battery module, which solve the problems that the influence of display factors on the battery health cannot be comprehensively and accurately reflected and the detection accuracy is low.
(II) technical scheme
In order to achieve the above purpose, the invention is realized by the following technical scheme: the system is used for determining the residual life and the health state of the retired battery module, and comprises a data acquisition module, a data preprocessing module, a characteristic extraction module, a residual life model building module, a residual life prediction module and a result output and health state evaluation module; the data acquisition module comprises a current sensor, a voltage sensor and a temperature sensor; the data preprocessing module cleans and preprocesses the collected original data; the feature extraction module extracts useful feature parameters from the preprocessed data; the residual life model building module builds a residual life prediction model by using the extracted characteristic parameters; the residual life prediction module predicts the residual life of the battery module by using a trained model; and the result output and health state evaluation module generates an indication according to the predicted result and generates an evaluation of the health state according to the result.
Preferably, the current sensor is used for monitoring the charge and discharge current of the retired battery module and outputting a current signal, the voltage sensor is used for detecting the voltage of the battery module and outputting a voltage signal, and the temperature sensor is used for detecting the temperature of the battery module and outputting a temperature signal.
Preferably, the data preprocessing module performs operations of removing abnormal values, smoothing data and calibrating the sensor according to the received data transmitted by the data acquisition module.
Preferably, the remaining life model predicts the remaining life by modeling the decay of battery life mathematical model.
Preferably, the residual life prediction module calculates through the established residual life model in combination with historical data and an algorithm to obtain the predicted residual life of the retired battery module, so that the data is transmitted to the result output and health state evaluation module.
Preferably, the result output and health state evaluation module generates an indication according to the prediction result transmitted by the remaining life prediction module.
The method for determining the residual life and the health state of the retired battery module is characterized by comprising the following specific steps:
s1, preprocessing training data by using a phase space reconstruction method, mapping two-dimensional data to a high-dimensional space, and performing offline training on a residual life model;
s2, taking the residual life model equation as an observation equation of a filter to obtain one-step prediction output;
s3, outputting capacity output of the double-filter algorithm and parameters of the residual life model to a feature extraction module to realize online training and dynamic parameter updating of the residual life model;
s4, parameter identification is carried out on the basis of the existing battery model, and the current battery health state is predicted through the change of corresponding parameters.
And S5.1, collecting large-scale performance data of the battery module, including parameters such as voltage, current, temperature and the like, and marking the parameters as different working states and health degrees. Marking the data according to the actual running condition of the battery module, and dividing the data into different health states (such as normal, damaged and retired) and corresponding residual lives (possibly represented by cycle times);
s5.2, deep neural network design: a deep neural network model is designed, and a cyclic neural network (RNN) is selected for the task of feature extraction and classification. The input of the model is the performance data of the battery module, and the output is the state of health classification and prediction of the remaining life of the battery.
S5.3, data preprocessing: preprocessing the acquired data, including normalization, noise reduction and data enhancement, to improve the robustness of the model.
S5.4, extracting characteristics: advanced features of the battery module data are extracted through the deep neural network model to capture hidden features which are not easy to observe, and the state of the battery can be described more accurately.
S5.5 classification and prediction: and inputting the extracted features into a deep neural network to perform health state classification and residual life prediction. This may include multi-class classification, including normal, damaged, and retired, as well as continuous numerical prediction, including estimation of remaining life.
Step a, having a series of battery performance data and their remaining life labels, denoted (x_1, x_2,) x_t) and (t_1, t_2, t_t), where x_i is battery performance data and t_i is the corresponding remaining life and state of health classification.
Step b, capturing a time series relation by using the RNN. The hidden state ht and output yt of the RNN can be derived by the following formula:
ht=σ(W ih *xt+W hh *ht-1)
where σ is the activation function, W ih And W is hh Is a weight matrix.
yt=W ho *ht
Wherein W is ho Is the output layer weight matrix.
The output layer may be a fully connected layer with 3 neurons, representing probabilities of three health states (normal, damaged, retired), respectively.
For continuous prediction of remaining life, the output layer may be a fully connected layer with 1 neuron, outputting an estimate of remaining life.
Step C, assuming that there is one RNN processed time series data x= { x_1, x_2,..x_t }, the goal is to predict both the remaining life T and the state of health class C of the battery.
Step d, the goal of the multitasking learning is to train the network together with the remaining life estimate and the health classification as two tasks. Here, two loss functions are introduced, one for residual life estimation and the other for health status classification.
Residual life estimation: for the remaining life estimation, the hidden state ht of the RNN is used to generate an estimated value. First, a fully connected layer with 1 neuron, denoted fT, is defined to generate a continuous prediction of remaining life:
T^t=fT(ht)
the Mean Square Error (MSE) loss is used in the training process to measure the difference between the actual remaining life Ti and the predicted value T≡i:
classification of health status:
for health status classification, three health status are generated using the output layer of the fully connected layer: probability of normal, damaged, retired, define a fully connected layer with 3 neurons, denoted fc, using softmax activation function to generate probability distribution for health status classification:
where c may be 0, 1 or 2, respectively, indicating normal, impaired and retired health status.
During training, cross entropy loss can be used to measure the difference of the actual health status label C from the predicted probability distribution:
wherein C is i,c Classifying different health states;
multiplexing loss:
in the multitasking learning, the loss of the remaining life estimate and the health status classification are comprehensively considered, and the total loss Lmulti is defined, which can be a weighted sum of the two losses:
Lmulti=αLT+βLC
where α and β are the weights of the losses, which can be adjusted according to the relative importance of the task.
Step e, training process: the model parameters θ are updated by minimizing the loss function with gradient descent:
where α is the learning rate.
Model evaluation: metrics such as MSE may be calculated using validation set or test set data to evaluate model performance.
S5.6, training a deep neural network model by using a large amount of marking data, and performing super-parameter tuning to ensure optimal performance.
S5.7, integrating the deep learning model with the residual life model parameter output characteristic extraction module in the S3 so as to realize online training and dynamic parameter updating. This allows the model to continuously adapt to the state change of the battery module over time.
S5.8 combines the filter output of S2 with the results of the deep learning model to comprehensively evaluate the state of health and remaining life of the battery, providing a more reliable prediction.
(III) beneficial effects
The invention provides a method and a system for determining the residual life and the health state of a retired battery module.
The beneficial effects are as follows:
1. the invention can monitor the performance and state change of the battery module in real time through iterative detection, can find problems such as battery capacity attenuation, internal resistance increase, temperature rise and the like in time through periodical detection, and other potential faults or abnormal conditions, can accumulate more data samples through repeated iterative detection, is beneficial to establishing a more accurate residual life prediction model, and can be used by staff to more accurately predict the life of the battery module and take corresponding maintenance and management measures in advance.
2. According to the invention, through parameter identification, a battery model can be fitted according to actual data, so that the behavior and performance characteristics of the retired battery can be more accurately described, the accuracy of predicting the battery health state can be improved, the monitoring result is more reliable, a worker can timely find the change trend of the retired battery health state, and when the parameter exceeds a preset range or abnormal change is displayed, the worker can warn in advance and take appropriate measures, so that serious consequences of battery faults or performance losses are avoided.
3. The deep learning model is capable of automatically learning and extracting useful features in the data without the need for manual design feature engineering. The system may be aided in better understanding of battery performance data, including complex time series information, to improve accurate predictions of battery state of health and remaining life. The deep learning classification model is introduced to remarkably improve the accuracy of the classification of the health state. The deep learning model can process a large amount of data and complex modes, so that the classification of health states such as normal, damaged, retired and the like is more reliable. The deep learning model can realize continuous prediction of the remaining life of the battery by learning a change pattern in the time series data. This allows for a more accurate estimation of when the battery may fail, helping to take maintenance action in advance, reducing unnecessary downtime and maintenance costs. The deep learning model may dynamically adapt to changing battery performance data. They can automatically adjust the model parameters to accommodate the new data so that the system can monitor battery health and remaining life in real time without the need to manually adjust the model. The deep learning technique can process large-scale data, and thus can be easily extended to a large-sized battery module management system. This helps to manage a large number of battery modules, improving the expandability and efficiency of the entire battery system.
Drawings
FIG. 1 is a flow chart of the system prediction of the present invention.
Detailed Description
The following description of the embodiments of the present invention will be made clearly and completely with reference to the accompanying drawings, in which it is apparent that the embodiments described are only some embodiments of the present invention, but not all embodiments. All other embodiments, which can be made by those skilled in the art based on the embodiments of the invention without making any inventive effort, are intended to be within the scope of the invention.
Embodiment one:
as shown in fig. 1, an embodiment of the present invention provides a system for determining a remaining life and a health status of a retired battery module, including a data acquisition module, a data preprocessing module, a feature extraction module, a remaining life model building module, a remaining life prediction module, and a result output and health status evaluation module; the data acquisition module comprises a current sensor, a voltage sensor and a temperature sensor; the data preprocessing module cleans and preprocesses the collected original data; the feature extraction module extracts useful feature parameters from the preprocessed data; the residual life model building module builds a residual life prediction model by using the extracted characteristic parameters; the residual life prediction module predicts the residual life of the battery module by using the trained model; the result output and health state evaluation module generates an indication according to the predicted result and generates an evaluation of health state according to the result.
Further, the current sensor is used for monitoring the charge and discharge current of the retired battery module, outputting a current signal, the voltage sensor is used for detecting the voltage of the battery module, outputting a voltage signal, and the temperature sensor is used for detecting the temperature of the battery module and outputting a temperature signal.
Further, the data preprocessing module performs operations of removing abnormal values, smoothing data and calibrating the sensor according to the received data transmitted by the data acquisition module.
Further, the remaining life model predicts the remaining life by modeling the decay of battery life mathematical model.
Further, the residual life prediction module obtains the predicted residual life of the retired battery module through the established residual life model and the combination of historical data and algorithm calculation, and accordingly data are transmitted to the result output and health state evaluation module.
Further, the result output and health state evaluation module generates an indication according to the prediction result transmitted by the remaining life prediction module.
The method for determining the residual life and the health state of the retired battery module is characterized by comprising the following specific steps:
s1, preprocessing training data by using a phase space reconstruction method, mapping two-dimensional data to a high-dimensional space, and performing offline training on a residual life model;
s2, taking the residual life model equation as an observation equation of a filter to obtain one-step prediction output;
s3, outputting capacity output of the double-filter algorithm and parameters of the residual life model to a feature extraction module to realize online training and dynamic parameter updating of the residual life model;
s4, parameter identification is carried out on the basis of the existing battery model, and the current battery health state is predicted through the change of corresponding parameters.
The remaining life of the retired battery module can be obtained by selecting retired battery capacity attenuation empirical data:
C k+1 =δ c C k +β 1 exp(-β 2 /Δt k ),
wherein C is k Is the battery capacity at k cycles, delta c Is coulombic efficiency, t k Is the time interval k to k+1, beta 1 And beta 2 Is a parameter to be determined;
therefore, a worker can obtain prediction data of the residual life of the retired battery module according to the data in the built model;
the formula is adopted: r is R k+1 =R k +r k 、y k =OCV(Z k )-R k i k +e k ,
Wherein: r is the internal resistance of the battery; r is (r) k Representing the change in internal resistance caused by noise; y is k Estimating a voltage for the battery; i.e k To estimate the current; z is Z k Is battery SOC, used for linking the relation of battery SOC-OCV; e, e k And representing errors of the battery model, thereby obtaining the residual life and the health state of the retired battery module with higher accuracy according to the final data.
And S5.1, collecting large-scale performance data of the battery module, including parameters such as voltage, current, temperature and the like, and marking the parameters as different working states and health degrees. Marking the data according to the actual running condition of the battery module, and dividing the data into different health states (such as normal, damaged and retired) and corresponding residual lives (possibly represented by cycle times);
s5.2, deep neural network design: a deep neural network model is designed, and a cyclic neural network (RNN) is selected for the task of feature extraction and classification. The input of the model is the performance data of the battery module, and the output is the state of health classification and prediction of the remaining life of the battery.
S5.3, data preprocessing: preprocessing the acquired data, including normalization, noise reduction and data enhancement, to improve the robustness of the model.
S5.4, extracting characteristics: advanced features of the battery module data are extracted through the deep neural network model to capture hidden features which are not easy to observe, and the state of the battery can be described more accurately.
S5.5 classification and prediction: and inputting the extracted features into a deep neural network to perform health state classification and residual life prediction. This may include multi-class classification, including normal, damaged, and retired, as well as continuous numerical prediction, including estimation of remaining life.
Step a, having a series of battery performance data and their remaining life labels, denoted (x_1, x_2,) x_t) and (t_1, t_2, t_t), where x_i is battery performance data and t_i is the corresponding remaining life and state of health classification.
Step b, capturing a time series relation by using the RNN. The hidden state ht and output yt of the RNN can be derived by the following formula:
ht=σ(W ih *xt+W hh *ht-1)
where σ is the activation function, W ih And W is hh Is a weight matrix.
yt=W ho *ht
Wherein W is ho Is the output layer weight matrix.
The output layer may be a fully connected layer with 3 neurons, representing probabilities of three health states (normal, damaged, retired), respectively.
For continuous prediction of remaining life, the output layer may be a fully connected layer with 1 neuron, outputting an estimate of remaining life.
Step C, assuming that there is one RNN processed time series data x= { x_1, x_2,..x_t }, the goal is to predict both the remaining life T and the state of health class C of the battery.
Step d, the goal of the multitasking learning is to train the network together with the remaining life estimate and the health classification as two tasks. Here, two loss functions are introduced, one for residual life estimation and the other for health status classification.
Residual life estimation: for the remaining life estimation, the hidden state ht of the RNN is used to generate an estimated value. First, a fully connected layer with 1 neuron, denoted fT, is defined to generate a continuous prediction of remaining life:
T^t=fT(ht)
the Mean Square Error (MSE) loss is used in the training process to measure the difference between the actual remaining life T i and the predicted value T≡i:
classification of health status:
for health status classification, three health status are generated using the output layer of the fully connected layer: probability of normal, damaged, retired, define a fully connected layer with 3 neurons, denoted fc, using softmax activation function to generate probability distribution for health status classification:
where c may be 0, 1 or 2, respectively, indicating normal, impaired and retired health status.
During training, cross entropy loss can be used to measure the difference of the actual health status label C from the predicted probability distribution:
wherein C is i,c Classifying different health states;
multiplexing loss:
in the multitasking learning, the loss of the remaining life estimate and the health status classification are comprehensively considered, and the total loss Lmulti is defined, which can be a weighted sum of the two losses:
Lmulti=αLT+βLC
where α and β are the weights of the losses, which can be adjusted according to the relative importance of the task.
Step e, training process: the model parameters θ are updated by minimizing the loss function with gradient descent:
where α is the learning rate.
Model evaluation: metrics such as MSE may be calculated using validation set or test set data to evaluate model performance.
S5.6, training a deep neural network model by using a large amount of marking data, and performing super-parameter tuning to ensure optimal performance.
S5.7, integrating the deep learning model with the residual life model parameter output characteristic extraction module in the S3 so as to realize online training and dynamic parameter updating. This allows the model to continuously adapt to the state change of the battery module over time.
S5.8 combines the filter output of S2 with the results of the deep learning model to comprehensively evaluate the state of health and remaining life of the battery, providing a more reliable prediction.
Although embodiments of the present invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made therein without departing from the principles and spirit of the invention, the scope of which is defined in the appended claims and their equivalents.
Claims (7)
1. A system for determining retired battery module remaining life and health status, its characterized in that: the device comprises a data preprocessing module, a feature extraction module, a residual life model building module, a residual life prediction module and a result output and health state evaluation module;
the data acquisition module comprises a current sensor, a voltage sensor and a temperature sensor;
the data preprocessing module cleans and preprocesses the collected original data;
the feature extraction module extracts useful feature parameters from the preprocessed data;
the residual life model building module builds a residual life prediction model by using the extracted characteristic parameters;
the residual life prediction module predicts the residual life of the battery module by using a trained model;
and the result output and health state evaluation module generates an indication according to the predicted result and generates an evaluation of the health state according to the result.
2. The system for determining the remaining life and health of an retired battery module according to claim 1, wherein: the current sensor is used for monitoring the charge and discharge current of the retired battery module and outputting a current signal, the voltage sensor is used for detecting the voltage of the battery module and outputting a voltage signal, and the temperature sensor is used for detecting the temperature of the battery module and outputting a temperature signal.
3. The system for determining the remaining life and health of an retired battery module according to claim 1, wherein: the data preprocessing module performs operations of removing abnormal values, smoothing data and calibrating the sensor according to the received data transmitted by the data acquisition module.
4. The system for determining the remaining life and health of an retired battery module according to claim 1, wherein: the remaining life model predicts the remaining life by modeling the decay of battery life mathematical model.
5. The system for determining the remaining life and health of an retired battery module according to claim 1, wherein: and the residual life prediction module obtains the predicted residual life of the retired battery module through the established residual life model and the combination of historical data and algorithm calculation, so that the data is transmitted to the result output and health state evaluation module.
6. The method and system for determining the remaining life and health of an retired battery module according to claim 1, wherein: and the result output and health state evaluation module generates an indication according to the prediction result transmitted by the residual life prediction module.
7. The method for determining the remaining life and health of an out-of-service battery module according to any one of claims 1 to 5, comprising the specific steps of:
s1, preprocessing training data by using a phase space reconstruction method, mapping two-dimensional data to a high-dimensional space, and performing offline training on a residual life model;
s2, taking the residual life model equation as an observation equation of a filter to obtain one-step prediction output;
s3, outputting capacity output of the double-filter algorithm and parameters of the residual life model to a feature extraction module to realize online training and dynamic parameter updating of the residual life model;
s4, parameter identification is carried out on the basis of the existing battery model, and the current battery health state is predicted through the change of corresponding parameters;
s5, feature extraction and classification based on deep learning;
s5.1, collecting large-scale battery module performance data, including voltage, current and temperature parameters, marking the data into different working states and health degrees, marking the data according to actual running conditions of the battery module, and dividing the data into different health states and corresponding residual lives: expressed in terms of cycle number;
s5.2, designing a deep neural network model, selecting a cyclic neural network (RNN) for characteristic extraction and classification tasks, wherein the input of the model is performance data of a battery module, and the output is the state of health classification and prediction of the residual life of the battery;
s5.3, preprocessing the acquired data, including normalization, noise reduction and data enhancement, so as to improve the robustness of the model;
s5.4, extracting advanced features of the battery module data through a deep neural network model to capture hidden features which are not easy to observe, so that the state of the battery can be described more accurately;
s5.5, inputting the extracted features into a deep neural network, and carrying out health state classification and residual life prediction, including multi-category classification including normal, damaged and retired, and continuous numerical prediction, including estimation of residual life;
step a, having a series of battery performance data and their remaining life labels, denoted (x_1, x_2,., x_t) and (t_1, t_2,., t_t), wherein x_i is battery performance data and t_i is the corresponding remaining life and state of health classification;
step b, capturing a time sequence relation by using the RNN, wherein the hidden state ht and the output yt of the RNN can be deduced by the following formula:
ht=σ(W ih *xt+W hh *ht-1)
where σ is the activation function, W ih And W is hh Is a weight matrix;
yt=W ho *ht
wherein W is ho Is an output layer weight matrix;
the output layer may be a fully connected layer with 3 neurons, representing probabilities of three health states (normal, damaged, retired), respectively;
for continuous prediction of remaining life, the output layer may be a fully connected layer with 1 neuron, outputting an estimate of remaining life;
step C, assuming that there is one RNN processed time series data x= { x_1, x_2,., x_t }, the goal is to predict the remaining life T and the state of health class C of the battery simultaneously;
step d, the goal of the multitasking is to train the network with the remaining life estimate and the health status classification as two tasks, where two loss functions are introduced, one for the remaining life estimate and the other for the health status classification;
residual life estimation: for the remaining life estimation, the hidden state ht of RNN is used to generate an estimated value, first, a fully connected layer with 1 neuron, denoted as fT, is defined to generate a continuous predicted value of remaining life:
T^t=fT(ht)
the Mean Square Error (MSE) loss is used in the training process to measure the difference between the actual remaining life Ti and the predicted value T≡i:
classification of health status:
for health status classification, three health status are generated using the output layer of the fully connected layer: probability of normal, damaged, retired, define a fully connected layer with 3 neurons, denoted fc, using softmax activation function to generate probability distribution for health status classification:
wherein c may be 0, 1 or 2, respectively, representing normal, impaired and retired health status;
during training, cross entropy loss can be used to measure the difference of the actual health status label C from the predicted probability distribution:
wherein C is i,c Classifying different health states;
multiplexing loss:
in the multitasking learning, the loss of the remaining life estimate and the health status classification are comprehensively considered, and the total loss Lmulti is defined, which can be a weighted sum of the two losses:
Lmulti=αLT+βLC
where α and β are the weights lost and can be adjusted according to the relative importance of the task;
step e, training process: the model parameters θ are updated by minimizing the loss function with gradient descent:
wherein α is the learning rate;
model evaluation: using the validation set or test set data to evaluate model performance, indices such as MSE can be calculated;
s5.6, training a deep neural network model by using a large amount of marking data, and performing super-parameter tuning to ensure the optimal performance;
s5.7, integrating the deep learning model with the residual life model parameter output characteristic extraction module in the S3 so as to realize online training and dynamic parameter updating; this allows the model to continuously adapt to the state changes of the battery module over time;
s5.8 combines the filter output of S2 with the results of the deep learning model to comprehensively evaluate the state of health and remaining life of the battery, providing a more reliable prediction.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202311465615.7A CN117471346A (en) | 2023-11-06 | 2023-11-06 | Method and system for determining remaining life and health status of retired battery module |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202311465615.7A CN117471346A (en) | 2023-11-06 | 2023-11-06 | Method and system for determining remaining life and health status of retired battery module |
Publications (1)
Publication Number | Publication Date |
---|---|
CN117471346A true CN117471346A (en) | 2024-01-30 |
Family
ID=89637580
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN202311465615.7A Pending CN117471346A (en) | 2023-11-06 | 2023-11-06 | Method and system for determining remaining life and health status of retired battery module |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN117471346A (en) |
Cited By (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN118169582A (en) * | 2024-05-15 | 2024-06-11 | 东方旭能(山东)科技发展有限公司 | Lithium ion battery health state and residual life prediction method |
-
2023
- 2023-11-06 CN CN202311465615.7A patent/CN117471346A/en active Pending
Cited By (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN118169582A (en) * | 2024-05-15 | 2024-06-11 | 东方旭能(山东)科技发展有限公司 | Lithium ion battery health state and residual life prediction method |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN112731159B (en) | Method for pre-judging and positioning battery faults of battery compartment of energy storage power station | |
CN111624494B (en) | Battery analysis method and system based on electrochemical parameters | |
EP2869153A1 (en) | Health management system, fault diagnosis system, health management method, and fault diagnosis method | |
KR101948604B1 (en) | Method and device for equipment health monitoring based on sensor clustering | |
CN102062832B (en) | Power system online prewarning method based on microdisturbance signal low-frequency oscillation mode identification | |
WO2002057856A2 (en) | Adaptive modeling of changed states in predictive condition monitoring | |
CN111046582A (en) | Nuclear power station diesel generating set coil temperature early warning method and system | |
Mosallam et al. | Component based data-driven prognostics for complex systems: Methodology and applications | |
CN117471346A (en) | Method and system for determining remaining life and health status of retired battery module | |
CN113447828A (en) | Lithium battery temperature estimation method and system based on Bayesian neural network | |
CN114462259B (en) | SOC-based battery remaining life prediction method and system, automobile and medium | |
CN117074965B (en) | Lithium ion battery remaining life prediction method and system | |
Yao et al. | Fault identification of lithium-ion battery pack for electric vehicle based on ga optimized ELM neural network | |
CN117330987B (en) | Method, system, medium and apparatus for time-based battery state of health assessment | |
CN114036647A (en) | Power battery safety risk assessment method based on real vehicle data | |
CN117553840A (en) | Instrument based on intelligent management and system thereof | |
CN117312972A (en) | Method for identifying health state of scraper conveyor speed reducer | |
CN117630672A (en) | Method and device for predicting a device battery of a diagnostic device using a converter model | |
CN116901707A (en) | Power battery pack fault early warning method, system and vehicle | |
Mosallam et al. | Integrated bayesian framework for remaining useful life prediction | |
CN115310693A (en) | Comprehensive energy data cleaning method, system, equipment and medium | |
RU2735296C1 (en) | Method for multi-parameter monitoring of complex electrotechnical objects state | |
CN112464146B (en) | Key subsystem based on historical telemetering data and single-machine correlation health baseline construction method | |
Degla et al. | An Enhanced Neural Network Application for a Lithium-ion Battery Pack State-of-Health Estimator | |
CN117272844B (en) | Method and system for predicting service life of distribution board |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
PB01 | Publication | ||
PB01 | Publication | ||
SE01 | Entry into force of request for substantive examination | ||
SE01 | Entry into force of request for substantive examination |